High-dimensional Statistical Methods for Biomedical Applications
Speaker:
Majnu John, PhD
Assistant Professor, Department of Psychiatry,
Zucker School of Medicine, Northwell/Hofstra University, Hempstead, NY
Associate Investigator, Feinstein Institute of Medical Research,
Northwell Health System, Manhasset, NY
Abstract:
High-dimensional biomedical datasets, such as genome-wide association studies and complex clinical or imaging data, present significant challenges for traditional statistical methods. In this talk, Dr. John will discuss modern high-dimensional statistical and machine-learning approaches that improve inference, prediction, and causal interpretation in biomedical research. A central focus will be on enhancing propensity-score–based adjustments for genetic association studies, where confounding from population structure, environmental factors, and correlated covariates can obscure true genetic effects. He will outline the motivation behind propensity score methodology in genomics and explain why conventional approaches struggle when the number of covariates greatly exceeds sample size. Recent advances, including regularization techniques, non-convex penalties, and data-adaptive machine-learning tools, have shown to yield more stable and interpretable models in high-dimensional settings. Using simulations and real genetic datasets, he will demonstrate how these methods improve bias reduction, variance control, and the reproducibility of association signals. Finally, the talk will highlight broader applications of these techniques in psychiatry, neuroscience, and other biomedical domains where complex, high-dimensional data require flexible analytical frameworks.
Dr. Majnu John earned his B.S. in Mathematics from Cochin University in India and completed his M.S. in Mathematics at the University of New Orleans. He received his Ph.D. in Applied Mathematics and Statistics from Johns Hopkins University in 2005, where his work focused on statistical methodology and computation. He went on to hold research positions at the Johns Hopkins School of Medicine and the Children’s Hospital of Philadelphia before joining Weill Cornell Medical College as an Instructor in Biostatistics. In 2010, Dr. John moved to the Feinstein Institute for Medical Research at Northwell Health, where he has built a multidisciplinary program in biostatistics, statistical learning, and biomedical data analysis. His research spans high-dimensional statistical methods, machine learning, dynamic time-series modeling, and applications in genetics, neuroscience, and psychiatric research. Dr. John is also an Adjunct Professor in the Department of Biomedical Engineering at NYU Tandon, where he contributes to teaching and collaborative research at the interface of statistics and biomedical engineering.
